Many representative agent macro models suffer from parameter identification failure. These issues extend to the structural parameters of the standard incomplete markets model of Aiyagari (1994), the workhorse heterogeneous agent model for household and firm behavior. Similar as in representative agent models, the firm-side parameters can display weak or failure of identification. Strong identification of these parameters can, however, be obtained by using a transformation of the data that amplifies the effect of the parameters. The household preference parameter can, nonetheless, not be identified using aggregate data alone, but it can be identified from the shape of the wealth distribution. This is in contrast with representative agent models, where the identification of this preference parameter comes from time variation in aggregate data. Using these results, a two-step estimation procedure is proposed, which is analogous to the manner of identification. Its performance is confirmed in a Monte Carlo simulation and applied to US data.